CN104704855B - For reducing the system and method for the delay in virtual low system for electrical teaching based on transposer - Google Patents

Abstract

The reduction system that postpones in virtual bass processing system performs harmonic transposition to generate the transposition data of the harmonic wave representing audio signal to the low frequency component of audio signal.This system uses the fundamental transposition factor more than 2, and in response to the frequency domain value determined by the asymmetric analysis window of use and the positive-going transition section of synthesis window and transformation by reciprocal direction section to generate harmonic wave.Virtual bass signal is combined with the wideband audio signal through time delay by this system by the analysis filterbank with the nyquist filter that filter coefficient blocks.When combining with audio signal, virtual bass signal can the delayed wideband audio signal through time delay to reduce the delay caused by harmonic transposition further.This virtual bass input signal can directly CQMF analysis filterbank from hybrid filter-bank section above be route, to avoid the time delay being associated with nyquist filter group.

Description

For reducing the system and method for the delay in virtual low system for electrical teaching based on transposer

Cross-Reference to Related Applications

This application claims in U.S. Provisional Patent Application No. that on October 15th, 2012 submits to The priority of 13/652,023, it is merged into herein with full content by quoting.

Technical field

One or more of embodiments usually relate to based on the Audio Signal Processing converted, and more More particularly to the delay reduced in virtual bass synthesis system based on transposer.

Background technology

Bass synthesis refers to add component to strengthen the bass of institute's perception to the low-frequency range of signal.? Among these methods, subwoofer synthetic technology creates the low frequency component of the existing partial less than signal, with The low-limit frequency scope that extension and raising are present in multi-object audio content.Another kind of method uses virtual Pitch algorithm, this virtual pitch algorithm (such as, is broadcast by little speaker from inaudible bass range The bass of the low pitch put) generate audible harmonic wave, therefore make harmonic wave and finally also make pitch listen Must see to improve bass response.

The synthesis of virtual bass is virtual pitch method, and it improves when can not physically reproduce low side bass The perception level of the bass content in audio frequency when playing audio frequency on the little speaker of frequency.The method based on " fundamental frequency of disappearance " psychoacoustics is observed below: even if when fundamental frequency and first harmonic self disappear, Human auditory system still is able to infer low pitch according to higher hamonic wave.Basic functional method is right The bass frequencies being present in audio frequency are analyzed and generate and contribute to the lower frequency that perception disappears Audible higher hamonic wave.Virtual bass be mainly characterized by comprising: that it passes through less than having little raising one's voice The higher hamonic wave of the frequency of the low-frequency roll-off (such as, less than 150Hz) of the device of device carries out synthesis and comes Strengthen perceived bass response on such devices.After energy adjusting, use multiple transposition factor Inaudible component of signal is replaced to higher audible frequency by (harmonic wave).Virtual bass synthesizes The perception bass of playback on headphones playback or full frequency band (full-range) speaker can also be increased. Figure 1A shows have the inaudible scope 10 of frequency component and higher than inaudible scope The Frequency and Amplitude spectrum of the audio signal of the audible scope of frequency component.In inaudible scope 10 The harmonic transposition of frequency component can generate the frequency through transposition in the part 11 of audible scope Rate component, it can strengthen the perception level of the bass content of audio signal during playback.Such humorous Ripple transposition can include each relevant frequency component to input audio signal apply multiple transpositions because of Son generates multiple harmonic waves of this component.

In some audio frequency processing system utilizing the low system for electrical teaching of traditional virtual, with frequency transposition function phase The time delay of association or delay can be the biggest for some is applied.Such as, delay is 1025 samplings Digital audio processing system can use the traditional virtual of the time delays adding other 3200 samplings Low system for electrical teaching.Assume sample frequency (f_s) be 48kHz, then this can cause total time delay more than 88 Millisecond.This retardation is the most problematic, and even prohibits for game and telecommunications are applied Only, wherein, the delay of about 100 milliseconds starts to become obvious in terms of audible signal lag.

In the low system for electrical teaching of traditional virtual use traditional transposer system for the time to frequency transformation Symmetrical time-domain window is used respectively with Analysis Block and the synthesis stage of frequency to time change.Figure 1B shows Go out relevant to the symmetrical window used in the low system for electrical teaching of traditional virtual as be known in the art The time delay of connection.Figure 1B shows the transposer i.e. being generated the second harmonic by second order transposer graphically The time delay forced.As shown in time graph 100, it is assumed that the time stride S of analysis window_A, The center of one of analysis window symmetrical on form is selected as time zero benchmark, and can be from analysis Time t in section 102₀Add newly inputted sampling 104.Time graph 110 shows transposer Time-stretching duality, wherein, t₀2 t being stretched in synthesis stage 112₀。

For the example process shown in Figure 1B, total analysis/synthesis chain time delay D_tsCan be below Equation 1 in be expressed as follows, wherein, L is transposer window size, S_AIt it is stride analysis time Or jump away from:

D_ts=L/2+2 (L/2-S_A)=3 L/2-2 S_A(equation 1)

In the audio frequency processing system organized based on HQMF (hybrid orthogonal mirror filter), extremely The input signal of CQMF (complex quad-rature-mirror wave filter) Analysis Block and synthesizing from CQMF The output signal of section typically each has identical sample frequency f_s, wherein, f_sIt is normally provided as 44.1kHz or 48kHz.Generally process due to system and be only from the first of 64 channel C QMF groups CQMF signal, so the input signal sampling rate processed for virtual bass can be f_s/64。 It should be noted that, it is also possible to use the CQMF size in addition to 64 passages.Due to usage factor 2 The fundamental transposition factor combination transposition function, so from traditional virtual bass processing system through transposition The sample frequency of output be 2 f_s/ 64, produce the factor 2 bandwidth expansion.In the transposer of combination, The fundamental transposition factor is that wherein source conversion frequency section (or frequency band) is mapped to according to man-to-man relation The factor of object transformation frequency band (or frequency band), i.e. be not related in source to target frequency section maps Interpolation or extraction.The fundamental transposition factor also controls the time stride of analysis window and the time of synthesis window Relation between stride.More particularly, generated time stride is multiplied by equal to stride analysis time and substantially changes Location factor.For L=64 and S_AFor the situation of=4, from based on 64 channel C QMF Time delay in the output sampling of system becomes:

D_ts={ 3 L/2-2 S_A64/2=2816 sampling (equation 2)

In addition to this time delay, it is also added with from two virtual bass output CQMF subband signals The time delay that Nyquist (Nyquist) filter bank analysis section processes.This time delay can be about 384 Individual sampling, thus for this example prior art traditional virtual bass processing system, give Total time delay of 2816+384=3200 sampling.

A solution for the delay forced by the low system for electrical teaching of traditional virtual is as by using Alternative components substitutes harmonic transposition device and changes the process Circuits System such as harmonic oscillator of reality.So And, this adds great amount of cost and complexity to system potentially, but also negatively affects audio frequency product Matter.

The theme discussed in background section should be simply because the most not referred It is regarded as prior art.Similarly, that mention in background section or and background section The problem that is associated of the theme problem that is not construed as previously having had realized that in prior art. Theme in background section is merely representative of different methods, and these different methods itself can also It it is invention.

Summary of the invention

Embodiment includes that the delay in virtual bass processing system reduces system, and it is to audio signal Low frequency component performs harmonic transposition to generate the transposition data representing harmonic wave.Harmonic transposition processes and uses greatly In the fundamental transposition factor of 2, and in response to by using asymmetric analysis window and the change of synthesis window The frequency domain value that the section of changing and inverse transformation section determine is to generate harmonic wave.Include blocking ptototype filter by use Nyquist analysis filterbank virtual bass signal is combined with the audio signal through time delay Generate the audio signal of enhancing.When combining with audio signal, virtual bass signal can be allowed stagnant Reduce by prolonging that harmonic transposition process causes with further by the audio signal limiting time section of time delay Late.

Embodiment includes by generating the low frequency component of input audio signal execution harmonic transposition Represent the method that the transposition data of harmonic wave reduce the delay in virtual bass generation system, wherein, humorous Ripple transposition uses the integer-valued fundamental transposition factor more than 2.Described method in response to by the time to frequency Rate territory transforming section and follow-up reverse frequency to time domain transforming section by the time to frequency domain transformation Frequency determined by asymmetric analysis window and synthesis window is used with the conversion of reverse frequency to time domain Thresholding generates harmonic wave.Input audio signal is subband CQMF (complex values quadrature mirror filter) Signal, and the sampling of input audio signal can be carried out pretreatment with generate represent low frequency component The audio frequency of threshold sampling.

In one embodiment, described method processes input sound by analysis filterbank or conversion Frequently signal is to provide group analysis subband signal or a frequency band according to low frequency component, use fundamental transposition because of Sub-B and transposition factor T are combined into subband signal or frequency band, and are filtered by synthesis Ripple device group or conversion come Treatment Analysis subband signal or frequency band to be combined into subband letter according to described one Number generate high fdrequency component.This represents the standard mode carrying out replacing, i.e. including conversion frequency section Perform forward FFT before the Nonlinear Processing mapped, then perform inverse FFT.Described Method can also include: in response to the transposition virtual bass signal of data genaration, and by virtual low Sound audio output signal applies one or more analysis filterbank by virtual bass signal and input Audio signal is combined to generate the audio signal strengthened, and wherein, analysis filterbank includes blocking Ptototype filter, it makes limited number of filter coefficient be removed.Described method can include again void Intend delayed, by by virtual low message relative to the predetermined amount of time of input audio signal of bass signal Number predetermined amount of time shorter compared with being delayed by processing time delay with virtual low system for electrical teaching is implicit Input audio signal combines, and the time lag generating the input sub-band sample combination included with time delay is empty Intend the audio signal of the enhancing of the sub-band sample that bass processes.

According to some embodiments, the fundamental transposition factor input audio signal in frequency domain is expanded to The degree that the value of the fundamental transposition factor matches is to produce the audio signal through transposition, and this fundamental transposition The factor can be the even-integral number value between 4 and 16.In one embodiment, to transposer CQMF The analysis filterbank that output subband carries out operating includes 8 passage nyquist filter group and 4 passages Nyquist filter group, and the ptototype filter coefficient of limited number of removal includes that 6 are Number.In another embodiment, input CQMF signal directly analyzes filter from CQMF above Ripple device group passage 0 outfan is route, therefore by follow-up nyquist filter group section bypass, and And therefore avoid relevant time delay.

The embodiment of described method can also include by the sample frequency to limit (when using analysis Between stride) generate windowing zero padding sampling come to input audio signal perform frequency domain over-sampling convert with life Become low frequency component.Owing to virtual bass signal can be allowed to delayed broadband input audio signal up to 20ms and there is not the obvious deterioration of the audio signal of enhancing, so when by virtual bass signal with When the input audio signal of time delay combines, predetermined amount of time can be to 1000 selected from 0 sampling The value of the scope of sampling.In one embodiment, asymmetric analysis window and synthesis window are configured Become the longer portion of analysis window is stretched towards past input sample, and make synthesis window Longer portion export sampling towards future and be stretched.

System that embodiment further relates to be configured to realize at least some method in said method or Equipment component.

Accompanying drawing explanation

In following accompanying drawing, identical reference is for referring to identical key element.Although below Figure depicts various example, but one or more is practiced without limitation to the example described in figure.

Figure 1A shows that the frequency component in known virtual bass processing system is from inaudible frequency The transposition of scope extremely audible frequency range.

Figure 1B shows symmetry that is well known in the prior art and that use in the low system for electrical teaching of traditional virtual The time delay that window is associated.

Fig. 2 is the general of the virtual bass processing system realizing postponing reduction process according to embodiment Include block diagram.

Fig. 3 A shows and mixes filter according to the pretreatment in the system based on HQMF of embodiment Ripple device group section.

Fig. 3 B shows that the previous Nyquist of the virtual bass processing system according to embodiment closes Become bank of filters section.

Fig. 3 C is the more detailed of the virtual bass processing system shown in the Fig. 2 according to embodiment Figure.

Fig. 4 be according to embodiment by virtual bass postpone that reduction process and system utilize main The block diagram of functional part.

Fig. 5 A show according to embodiment with not same order virtual using the fundamental transposition factor Bass postpones the first jumping of the reduction system form away from the time delay being associated.

Fig. 5 B show according to embodiment with not same order virtual using the fundamental transposition factor Bass postpones the second jumping of the reduction system form away from the time delay being associated.

Fig. 5 C is the example of the time response of the asymmetric window compared with some conventional symmetrical window Figure, and Fig. 5 D is the showing of frequency response of asymmetric window compared with some conventional symmetrical window Illustration.

Fig. 6 shows the use of the asymmetric window according to embodiment and by B rank fundamental transposition The time delay being associated that device is forced.

Fig. 7 A shows the asymmetric mapping window of the use according to embodiment and the fundamental transposition factor The virtual bass of not same order postpone the first jumping of reduction system away from the form of total retardation value.

Fig. 7 B shows the asymmetric mapping window of the use according to embodiment and the fundamental transposition factor The virtual bass of not same order postpone the second jumping of reduction system away from the form of total retardation value.

Fig. 8 shows including virtual bass generation system and postponing to reduce system according to embodiment The block diagram of the audio frequency processing system of system.

Detailed description of the invention

Describe for reducing the delay in virtual low system for electrical teaching based on transposer and algorithm time delay The embodiment of system and method.Such system and method utilizes the higher order fundamental transposition factor, low Postpone asymmetric mapping window, block Nyquist ptototype filter, relative to original audio signal How being bypassed in the virtual bass signal of time lag and hybrid filter-bank section previously Qwest's analysis filterbank.

Run through the present disclosure including claims, " to " signal or data perform operation The expression of (such as, being filtered signal or data, scaling, convert or apply gain) is used for Broadly represent to signal or data or to the processed version of signal or data (such as, to letter The version of the signal of preliminary filtering or pretreatment is had gone through before number performing operation) directly perform behaviour Make.Express " transposer " to hold for representing in a broad sense to compose for part or whole available input signal The pitch shift (pitch-shifting) of row real number value or complex values input signal or the calculation of time-stretching Method unit or device.Express " transposer ", " harmonic transposition device ", " phase vocoder ", " high frequency life Grow up to be a useful person " or " harmonic generation device " can be used interchangeably.Express " system " in broad sense Upper expression device, system or subsystem.Such as, it is achieved the subsystem of decoder can be referred to as decoding Device system, and include that the system of such subsystem (such as, generates X in response to multiple inputs The system of individual output signal, wherein subsystem generates M input, and other X-M input Receive from external source) decoder system can also be referred to as.Term " processor " is in a broad sense Represent and may be programmed to or otherwise can be configured to (such as, using software or firmware) to data (example As, audio or video or other view data) perform operation system or device.The example of processor Including field programmable gate array (or other configurable integrated circuit or chipsets), be programmed to and/ Or otherwise it is configured to the numeral that audio frequency or other voice data execution pipeline formulas are processed Signal processor, programmable universal processor or computer and programmable microprocessor or core Sheet or chipset.Express " audio process " and " audio treatment unit " to be used interchangeably And in a broad sense for representing the system being configured to voice data is processed.Audio treatment unit Example include but not limited to encoder (such as, code converter), decoder, vocoder, volume Decoder, pretreatment system, after-treatment system and bit stream processing system are (sometimes referred to as bit stream Handling implement).

Embodiment relates to reduce virtual bass time delay and without to existing virtual bass process portion The harmonic transposition device used in part such as virtual bass processing system carries out the system and method for material change. Can make in conjunction with the harmonic generation device (transposer) in audio codec (such as, decoder) Postpone to reduce the various aspects of system and method with virtual bass.Can be combined with other transposers or phase Position vocoder system is such as the time-stretching for audio signal or traditional phase of pitch shift Position vocoder uses virtual bass to postpone to reduce the various aspects of system and method.

The most as shown in Figure 1A, the virtual bass using harmonic transposition generates method and includes frequency Component exists to improve bass content to the transposition of audible frequency range from inaudible frequency range Such as by can not physically reproduce the little speaker of the lower frequency of disappearance in limited playback equipment Playback.Virtual bass postpones to reduce the embodiment of system and method and improves virtual bass generation side Method, this virtual bass generates method and represents the low frequency component execution harmonic transposition of audio signal to generate It is contemplated that the data through transposition of audible harmonic wave during Hui Fang, in response to the data genaration through transposition Virtual bass signal, and by by virtual bass signal and (through time delay) input audio signal group The incompatible audio signal generating enhancing.Generally, in the audio signal strengthened by can not physically reproduce During the playback of one or more speaker of low frequency component, in the audio signal of enhancing provides bass The perception level that improve held.

The two of the harmonic transposition utilization each low frequency component of use that method performs are generated by virtual bass Rank transposer and at least one higher order transposer (generally, three rank transposers and quadravalence transposer, with And alternatively, at least one other higher order transposer) combination transposition generate harmonic wave so that In response to by the common time to frequency domain transformation section (such as, by from the single time to frequency domain The coefficient of frequency that obtains of conversion performs other operations of phase place multiplication or phase place) frequency domain value next life of determining Become all harmonic waves, carry out after the above-mentioned common time to frequency domain transformation section common frequency to time Between territory conversion (it practice, above-mentioned common frequency to time domain conversion be divided into two less conversion To adapt to bandwidth and the sample frequency of the subband of CQMF framework).

Fig. 2 is that some the delay reductions that realize according to embodiment process or combine some delay reductions Process the block diagram of the virtual bass processing system used.In one embodiment, at virtual bass Reason system 200 obtains multiple complex values sub-band sample from so-called hybrid filter-bank, and (HQMF adopts Sample) as input 201 (input A).In one embodiment, before virtual bass processes Hybrid filter-bank original time domain audio input signal has been separated into such multiple hybrid subband 201 (they are discussed in further detail below), and they can cache by input buffer 206. Then, the Nyquist synthesis filter banks 208 that enters through cached is processed, and Nyquist closes Become bank of filters 208 to perform complex functionality and represent that low-frequency audio content is (such as, at 0Hz with reconstruct And between 375Hz) single complex values QMF (CQMF) territory signal 202 (signal C). In another embodiment, virtual low system for electrical teaching is by by the Nai Kui in hybrid filter-bank above This special bank of filters section bypass includes postponing the mechanism of saving.This allows the system to by inciting somebody to action CQMF passage 0 signal is directly fed to virtual bass module as input 203 (input B) and saves Save the time delay being associated with Nyquist analysis group (such as, 384 samplings).As shown in Figure 2, One of two inputs 202 or 203 are selected by the switch such as selector 204, and selected signal Including virtual bass input signal 205 (signal D), it is processed by transposer 209 further.

Transposer (or phase vocoder) is typically following combination: in time to frequency transformation or filtering It is non-linear section (performing phase place multiplication or phase-shifts) after device group, is frequency after non-linear section Rate is to time change or bank of filters.Thus, as shown in Figure 2, transposer 209 includes that the time is extremely Frequency transform part 210, non-linear section 212 and frequency are to time change 214.In transposer 209 Non-linear section 212 be amendment phase place and subband or transform component to signal apply some gains The process block of (amplitude) control signal.Then, the signal through replacing is cached by output buffer 216, And processed by Nyquist analysis filterbank 218 subsequently, Nyquist analysis filterbank 218 Perform virtual bass output CQMT signal decomposition is become the hybrid subband sampling with input signal 201 (HQMF) analytic function of corresponding subband.Input the undressed through time delay of a-signal To produce, version 2 20 is mixed with the output of nyquist filter group 218 includes that virtual bass exports Signal is plus the audio output signal 222 of the enhancing of the input signal through time delay.

Although embodiment can relate to nyquist filter group for some function such as synthesis 208 The uses processed with analysis 218 sections, it should be noted that other kinds of wave filter can also be used Group or frequency division or divide circuit and technology.In other embodiments, wave filter above-mentioned Group or frequency division or divide circuit and technology can not exist.

Fig. 3 A to Fig. 3 C is the more detailed figure of the virtual bass processing system shown in Fig. 2.Figure 3A shows pretreatment hybrid filter-bank section 300, i.e. be frequently not the one of virtual low system for electrical teaching Part but alternatively section before virtual low system for electrical teaching.Hybrid filter-bank can be CQMF The combination of group, wherein, a number of minimum CQMF band is by the nyquist filtering of predefined size Device group processes to improve the frequency resolution of low-frequency range.From Nyquist Analysis Block and remaining The combination of the low frequency sub-band sampling of CQMF passage is referred to as hybrid subband sampling or HQMF (mixing QMF) signal.As shown in fig. 3, time domain input signal 302 is input to 64 channel C QMF Analysis filterbank 304.In one embodiment, output, a CQMF of this bank of filters Passage 0 (being expressed as signal B) 306 is directly fed to the virtual bass module 330 of Fig. 3 C (should Signal is corresponding to the input B 203 of Fig. 2).It should be noted that, Nyquist analysis is filtered by signal B 306 Ripple device group 307 bypasses, and therefore avoids the time delay being associated.CQMF passage 0,1 and 2 are the most defeated Enter to multiple Nyquist analysis filterbank 307 to 309.From Nyquist analysis filterbank Output with remaining CQMF subband (3 to 63) produces hybrid subband sampling 0 to 76 (by table It is shown as signal A) 310.

As shown in the system 320 of Fig. 3 B, multiple complex values hybrid subband sampling (signal A) 322 It is input to Nyquist synthesis filter banks section 324.Assuming that the virtual bass module 330 of Fig. 3 C Being one of other modules in system, hybrid subband sampling (HQMF sampling) is operated by it. Therefore, the signal A 310 of Fig. 3 A can be in pretreatment before becoming the input A 322 of Fig. 3 B The process of other modules is experienced after bank of filters section 300.In a kind of example embodiment, first 8 hybrid subband are i.e. from low frequency 8 passage (8-ch) nyquist filter group 307 (its basis Sampling rate produces the signal bandwidth of substantially 344Hz to 375Hz) subband be processed.Due to how Nyquist filter group is contrary with CQMF group not to be downsampled, so nyquist filter combination Become step particularly simple, because it is only the sub-band sample of each CQMF (or HQMF) time slot Summation.After 8 minimum hybrid subband samplings are sued for peace by section 324, System reorganization CQMF passage 0 signal C 326, it becomes the input 332 of virtual bass module 330 of Fig. 3 C.

Fig. 3 C shows that some the delay reductions that realize according to embodiment process or combine some delays Reduction processes the virtual low system for electrical teaching used.The virtual bass module 330 of Fig. 3 C is by signal D 332 As input.In the embodiment that Nyquist analysis filterbank 307 above is bypassed, letter Number D 332 can be route from the signal B 306 of Fig. 3 A.In another embodiment, signal D 332 can be fed from the signal C 326 of the Nyquist Analysis Block 320 of Fig. 3 B.This two Planting in embodiment, the input signal of that is virtual bass module of signal D 332 is single complex values CQMF signal (that is, from the first passage (passage 0) of one group of CQMF subband signal).

In virtual bass is applied, can perform optionally dynamically to process function by dynamic processor 336 To change the dynamic of virtual bass input signal.Processor 336 may be used for reducing the level of weak bass And keep or strengthen strong bass, i.e. being used as expander.The program meets in bass range Deng the shape of sound profile (ELC), wherein, loudness contour is flatter in the frequency of louder signal, And signal for more weak loudness is more precipitous.Therefore, when generate harmonic wave with keep fundamental component with life Become harmonic wave between relative loudness time, more weak bass can be attenuated more by stronger bass.Dynamically The gain of processor 336 can be by (running) averaged energy signal that slides, i.e. a CQMF The moving average energy through (monophonic) version of lower mixing of band signal 332 controls.

For the embodiment of system 330, before input to Nonlinear Processing block 344, to (can Dynamic process) CQMF signal perform use window size L (include the zero of up to length N Fill), the first of forward FFT 340 and modulation function 342 add window function.Enforcement in the present invention In mode, window shape is asymmetrical.In another embodiment, transposer (includes parts 338 to 356) representing the phase vocoder improved, it uses the FFT identical with fundamental transposition device Analyze/synthesis chain use and be referred to as the interpolation technique of " combination transposition " with generate second order, three rank, Quadravalence and the harmonic wave (the transposition factor) of possible higher order.Generally, although infringement removes to a certain extent The quality of other harmonic waves beyond base order harmonics, but such combination transposition is saved and is calculated complexity Degree.In the case of not using combination transposition, at least positive-going transition or inverse transformation is required for different Transposition the factor and different.Nonlinear Processing block 344 uses integer to replace the factor, and it carries out certain of redundancy Individual phase estimation, phase unwrapping or phase locking techniques, these technology generally ought be used for a lot of standard Phase vocoder in time unstable and inaccuracy.In one embodiment, phase multiplier 344 Use the fundamental transposition factor B such as 8 higher than 2 or any other suitable value.

Transposer 338 to 356 uses over-sampling (that is, the zero padding in block 338 and 356 in a frequency domain Fill analysis window and synthesis window) improve pulse (percussion) sound, it is when at bass frequencies model Enclose middle use time be main.In the case of not carrying out such over-sampling, tap tum sound and very may be used At least some Pre echoes and rear echo puppet sound can be generated so that bass is fuzzy and unintelligible.Real in one Executing in mode, oversample factor F is selected as at least factor F=(B+1)/2, and wherein, B is basic The transposition factor (such as, B=8).This contributes to guaranteeing the transient sound suppression pre-echo for isolation With rear echo.

As shown in FIG. 3 C, transposer includes by phase multiplier circuit (Nonlinear Processing block 344) gain of every FFT frequency band of amplifier 346 application after and slope-compensation.This makes The overall gain of the different transposition factors can be set independently.Such as, gain can be configured to approximation Wait in some and ring profile (ELC).As approximation, can right by the frequency less than 400Hz Straight line on number scale suitably to be modeled ELC.In this case, although odd-order Harmonic wave (such as, three rank, five rank etc.) is critically important for the virtual bass effect of result, but due to Odd-order harmonics can be perceived as compared with even-order harmonics more ear-piercing sometimes, so odd-order harmonics Can be attenuated largely.Each signal through transposition can additionally have slope gain, i.e. The roll-off attenuation factor measured with such as every octave dB.Also at transform domain by amplifier 346 Every frequency band applies this decay.

In system based on non-mixed bank of filters, such as, in time domain system, use the letter of Fig. 3 A Numbers 302 as input, and transposer 338 to 356 will be directly to fully sampled speed (such as, 44.1kHz Or 48kHz) time-domain signal operate, then utilize the FFT size of substantially 4096 lines, To provide the suitable resolution of low frequency (bass) scope.But, in one embodiment, right CQMF passage 0 sub-band sample (the signal D 332 of system 330) performs all process.This passes through Signal interested is processed by transposer, i.e. (or maximum by threshold sampling Extraction) low-pass signal is processed to provide relative to normal some advantage such as joint processing practice Save computation complexity.Such as, by using quadravalence fundamental transposition device, virtual low system for electrical teaching passes through the factor 4 bandwidth extending input signal.Generally, it is not required that virtual low system for electrical teaching output bandwidth is higher than substantially The signal of 500Hz.This represents, carries a width of 375Hz (or f_s=48kHz) a CQMF lead to Road (passage 0) is the most suitable for the input of virtual bass, and above two CQMF passages are (logical Road 0 and 1) for virtual bass exports, there is enough bandwidth (at f_s750Hz under=48kHz). Using CQMF passage 0 as input, this system can use size 64 (4096/64) to replace 4096 FFT process complex values sampling, wherein, be reduced to 1/64 and be derived under CQMF group Decimation factor, compared with time domain input signal, this is also equal to the reduction of a CQMF subband signal Bandwidth.Due to intrinsic bandwidth expansion, so the output from transposer needs to be transformed into CQMF band 0 and 1.This can also calculate the composition CQMF two 16 line FFT with 0 and 1 Inverse FFT before by 64 line FFT being divided into 4 16 line FFT and followed by conversion CQMF ptototype filter response compensation in territory is carried out approx.Note, superincumbent example In, do not consider frequency domain over-sampling, become because it will increase forward by previously mentioned oversample factor Change size and inverse transformation size.In one application, can be in the module of virtual bass module 330 Split FFT spectrum in 348, and CQMF wave filter response benefit can be carried out by multiplexer 350 Repay.At other embodiments, before FFT segmentation module 348, can be to complete (on such as, 64 lines in the example of face) FFT spectrum carry out CQMF wave filter response compensate.

As shown in Fig. 3 C further, use N/B the transform size put and windowing subsequently With overlapping/addition step 356, use length of window L/B, will mend from the response of CQMF wave filter Repay the output input of block 350 to the modulation step 352 before inverse fft circuit 354.At this In bright embodiment, window shape is asymmetrical.Block 348 He can also be split at FFT Application modulation step 352 before CQMF wave filter response compensation block 350.From windowing and overlapping/ The output signal of added circuit 356 is two CQMF signals, including will be with the HQMF through time delay The virtual bass signal of signal A 364 mixing.But, the two signal is firstly the need of respectively by 8 Passage and 4 passage Nyquist analysis filterbank 360 are filtered and are met hybrid domain.In the present invention Embodiment in, Nyquist analysis filterbank 360 use block ptototype filter.From filter Ripple device group 360 HQMF output can by bandpass filtering, and in module 362 with through time delay Input component A 364 mix with produce strengthen audio frequency output HQMF signal 366.In one In embodiment, to the time delay of input A 364 of mixed zone mixed block 362 less than virtual low system for electrical teaching Time delay (if using signal B 306 as input, then deduct Nyquist and analyze time delay) is to include The virtual bass signal of time lag.

When as outlined above, when performing FFT segmentation, will not keep analyzing from CQMF Phase relation between the subband signal of group.In order to alleviate this situation, system in embodiments CQMF passage 1 was used by exp (-j pi/2) multiplication before Nyquist analysis block 360 by 330 The phase compensation of 358.The specific independent variable of phase compensation function 358 depends on before by Fig. 3 A The modulation scheme that uses of CQMF group 304, and can be different because of embodiment.Additionally, Compensating factor 358 can be moved in other process block and be absorbed.

Virtual bass postpones to reduce

As described in the background section, virtual bass processing system is drawn when processing input signal Enter some time delays.With reference to Figure 1B, the time delay of tradition transposer is (about transposer output sampling frequency rate Measured) D=3 L/2-2 S can be expressed as_A, wherein, L is transposer window size, S_A Be analyze stride or jump away from.As previously described, at L=64 and S_AIn the system of=4, transposition Total time delay of device and nyquist filter group analysis section can be about 3200 samplings.

In one embodiment, virtual bass processing system includes performing some steps to reduce with empty Intend bass and process the parts of the delay that content is associated.Fig. 4 be according to embodiment by virtual bass Postpone the block diagram of the main function components of reduction process and system utilization.As shown in Figure 40 0 of Fig. 4, Postpone reduction process to include using the higher order fundamental transposition factor 402, the asymmetric mapping window of low latency 404, the virtual bass signal 408 of Nyquist ptototype filter 406 and time lag is blocked.Figure Each functional part of 400 can be used alone or combine one or more in miscellaneous part Use to help to reduce the delay of virtual bass process content.Such as when in parts 402 to 408 When each is embodied as hardware component such as circuit, processor etc., Figure 40 0 can represent system.Example If the action being implemented as being performed by functional part when each in parts 402 to 408 is as by one Or more process perform computer implemented process time, this figure also may indicate that process.Alternative Ground, Figure 40 0 can represent some of them parts can so that ware circuit realizes miscellaneous part May be implemented as hybrid system and the method for performed method step.Parts 402 to 408 are permissible It is implemented as different individual components, or they can be combined in prolonging of one or more merging Reduce late in function.The composition of each parts of system 400 and being described in detail as follows of operation.

The higher order fundamental transposition factor

For the higher order fundamental transposition factor 402 of Fig. 4, can be by tradition transposer time delay equation D_ts={ 3 L/2-2 S_A64/2 (equation 2) be derived as shown in Equation 3:

D_ts={ (B+1) L/2-B S_A64/B (equation 3)

In equation 3, the fundamental transposition factor 2 of legacy system is by arbitrary integer fundamental transposition factor B Replace.Noting, equation 3 refers to have the output of the framework based on CQMF of 64 passages and adopts The time delay of sample.Can verify, for constant L and S_AFor, time delay reduces along with the increase of B. For according to embodiment virtual bass postpone reduction system, Fig. 5 A show with first jump away from The time delay being associated, Fig. 5 B shows and jumps away from the time delay being associated with second.The form of Fig. 5 A 1 shows for various window sizes (L=16 to 128) and the fundamental transposition factor (B=2 to 16) For jump away from S_AThe delay of=4.By contrast, the form 2 of Fig. 5 B shows for identical each Jump away from S for planting window size (L=16 to 128) and the fundamental transposition factor (B=2 to 16)_A=2 Delay.As seen in Fig. 5 A and Fig. 5 B, such as by by the fundamental transposition factor from 2 increase to 8, it is possible to achieve significantly postpone to reduce (such as, for L=64 and S_AThe nominal of=4 For situation, it is reduced to 2048 samplings from 2816 samplings).

With reference to Fig. 3 C, in the transposer 338 to 356 of combination, when generating the higher order transposition factor When T wherein T is more than B (T ＞ B), transposer source range is less than the transposer in analytic transformation spectrum Target zone.Target frequency section is produced by the interpolation of source frequency section.When using higher order fundamental transposition device Generate low order to replace because of the period of the day from 11 p.m. to 1 a.m, i.e. when T is less than B (T ＜ B), source range will be greater than target Scope, and target frequency section is by the extraction generation of source frequency section.But, similarly for situation T ＜ For B, when T is odd number, the source frequency segment index being derived as k=n B/T is typically not integer Wherein n is target frequency segment index, and therefore by the interpolation from two continuous print source frequency sections Derive target frequency section.

Virtual bass is processed by the exponent number increased of the fundamental transposition factor has certain association.First First, need to set up control with in the range of forcing transposer source range to be maintained at analytic transformation (that is, 0 to In the range of N-1).Secondly, compared with the system using the fundamental transposition factor 2, present two conjunctions Becoming conversion 354 sizes can be N/B rather than N/2, wherein, N is analytic transformation size.This represents, Synthesis window will by factor B rather than 2 be extracted and compose segmentation 348 together with wave filter response mend The gain vector repaying 350 also can be the most reduced.This is adding of the high value for B The result of bandwidth expansion；The frequency range that transposer output covers B CQMF band inherently is (false The input of fixed CQMF band), wherein, the most above two CQMF bands are synthesized, Thus save complexity.For fundamental transposition factor B=8 and frequency domain oversample factor F=4, Two synthesis transform size are N_S=F L/B=4 64/8=32, and synthesize mapping window 356 and only have L/B=64/8=8 tap.

The quality of the signal through replacing is controlled by the fundamental transposition factor, and for higher order transposition rank Decrease for number, but can be by using the analysis reduced to jump away from (increasing in time domain Over-sampling) improve.Additionally, in order to keep the quality tapping sound (transition), for higher The fundamental transposition factor, need increase frequency domain over-sampling exponent number.But, in time and frequency two The over-sampling added of aspect may increase the computation complexity of transposer.At a kind of embodiment In, compared with legacy system, analyze and jump away from being reduced 1/2nd.Substantially changing of factor B=8 Position device would be required to the frequency domain oversample factor of at least F=(B+1)/2=4.5.In one embodiment, The over-sampling (F=4) of this system usage factor 4, and when mapping window disappears when end comes to a point Value 0.5 the most notable.Therefore, in this embodiment, owing to increasing in time Over-sampling, is caused computation complexity to amount to and is increased by the factor 2.It should be noted that, increase time Between over-sampling with the time delay that slightly increases as cost, for L=64, B=8 and S_AFor=2, with 2176 sampling total delaies and terminate, as shown in the form 2 of Fig. 5 B.

Asymmetrical mapping window

Give the content shown in the form 1 of Fig. 5 A and Fig. 5 B and form 2, it can be assumed that reduce The obvious way of transposer time delay is to use shorter mapping window and the least analysis to become Change size and synthesis transform size.But, this generally with reduce intensive tone signal quality as cost, Because being produced the frequency resolution reduced by shorter mapping window.Have been found that and can pass through Positive-going transition section and inverse transformation section use asymmetric analysis window and synthesis window to realize transposer The more robust reduction of algorithm time delay.Thus, in one embodiment, low for Fig. 4 is prolonged For slow asymmetric conversion 404, postpone reduction system and use non-in positive-going transition section and inverse transformation section Symmetry analysis window and the synthesis window windowed segments 338 and 356 of Fig. 3 C (such as, be respectively).This Do not caused conversion time delay substantially to improve to have by " tail " towards history samples extended window The frequency response of the asymmetric window of limit for length.In the most more generally embodiment, analysis window The size of length and positive-going transition can be with the varying in size of the length of synthesis window and inverse transformation.

Fig. 5 C is the time of asymmetric window compared with peaceful (Hanning) window of the Chinese with conventional symmetrical The example plot of response.Fig. 5 C shows: as shown in curve chart 514 for a length of 64 Conduct sampling (x-axis) of Hanning window mouth and the function of signal amplitude (such as, in units of volt) Time response, as shown in curve chart 516 for a length of 41 Hanning window mouth conduct sampling (x-axis) and the time response of function of signal amplitude (such as, in units of volt), Yi Jichang The asymmetric window that degree is 64 and time delay is 40 (time delay Hanning window mouths equal to a length of 41) Time response curve chart 512.Fig. 5 D is asymmetric window compared with the Hanning window mouth of conventional symmetrical The exemplary graph of frequency response.Fig. 5 D shows: as shown in curve chart 524 for a length of The Hanning window mouth of 64 as the signal amplitude on normalized frequency (x-axis) and logarithmic scale (such as, The frequency response of function dB), as shown in curve 526 for a length of 41 Hanning window mouth Function as signal amplitude (such as, dB) on normalized frequency (x-axis) and logarithmic scale Frequency response and a length of 64 and time delay be the non-of 40 (equal to the Hanning window mouths of length 41) The frequency response curve 522 of symmetrical window.As seen in figure 5d, asymmetric window Main lobe (lobe) has between two Hanning window mouths of expression between the Hanning window mouth of these symmetries Frequency resolution or selective width.

In order to adapt to asymmetric window conversion process, compared with tradition realization, need partly to change to change Position device algorithm, it is considered to analyze/synthesize the conversion time delay D reduced of chain.Replace at legacy system E is passed through after positive-going transition and before inverse transformation^-jπkCarry out frequency modulation(PFM), asymmetric system require with Frequency modulation(PFM) 342 is carried out after lower analytic transformation:

M_A(k)=e^{-i·(2·π/N)(D/2-L+1)·k}, 0≤k ＜ N (equation 4)

This system also requires to be modulated before the segmentation of following synthesis FFT spectrum:

M_S(n)=e^{-i·(π/N·D·n)}, 0≤n ＜ N (equation 5)

In superincumbent equation 4 and equation 5, k and n is conversion frequency coefficient index respectively, and N is Analytic transformation size, i.e. N=FL, wherein, F is frequency domain oversample factor, and L is analysis window Size and D are conversion time delays.As pointed by Fig. 3 C, the modulation of equation 5 can also be answered Modem section 352 after FFT segmentation module 348 and response compensation process 350.

Fig. 6 shows the use of the asymmetric window according to embodiment and by B rank on form The time delay being associated that fundamental transposition device is forced.In the low system for electrical teaching of traditional virtual, B is generally positioned Become 2, but if combine higher order fundamental transposition factor treatment 402 to use asymmetric window treatments 404, then B can be greater than the integer value (such as, B=4,8 or 16) of 2.Time graph 600 shows Go out the time zero benchmark of the group delay as analysis window (approximating D/2).Analysis Block 602 from Time t₀Rise and add new sampling 604.Time graph 610 shows the time-stretching antithesis of transposer Property in the synthesis stage 612 of the sampling 614 of new time-stretching by t₀Mobile to time B t₀.Using In the case of asymmetric window window as shown in Fig. 5 (512) or Fig. 6, total analysis/synthesis chain Amount of delay is similar to: D/2+B (D/2-S_A)。

Can be realized by the circulation time displacement sampled by N/2 for wherein frequency domain modulation For symmetrical window situation, the calculating of equation 4 above and equation 5 can be the most respectively by dividing The circulation time of N-(D/2-(L-1)) (mod N) the individual sampling before analysis conversion shifts and (single) The circulation time displacement of N-D/2 sampling after synthesis conversion realizes.But, when by asymmetric When window combines with the higher order fundamental transposition factor such as B=8 and FFT segmentation section 348, synthesis Time shift after conversion can be (N-D/2)/B sampling, and it can be integer value.At this In the case of Zhong, the value rounded can serve as approximation.It addition, in order to save complexity, can will divide Analysis modulation is modulated as the synthesis merged be given by equation 6 with synthesis furnishing combination:

M_ASC(k)=e^{-i·(2·π/N)(D/2·(B+1)-L+1)·B)·k}, 0≤k ＜ N (equation 6)

Only when factor T that replaces is equal to B, the hybrid modulation of equation 6 just can be accurately.Other are replaced The factor, equation 6 also can be approximation.

Alternately, the synthesis that the modulation of equation 6 may be implemented as shown in equation 7 converts it After combination circulation time displacement:

$\begin{array}{cc}{f}_x\left(m\right)=g_x(S+m)& ,0\&le;m<N/B-S\\ f_x(N/B-S+m)=g_x\left(m\right)& ,0\&le;m<S\end{array}$ (equation 7)

In superincumbent equation 7, g_xM () is the time domain output from one of synthesis inverse transformation, f_x(m) It is the time series of displacement, and S is equal to:

Additionally, when ceil functionThe independent variable of (rounding into immediate integer) is not accurate During integer, equation 7 provides the warbled only approximation realized by equation 6 (itself can be Approximation).Shall also be noted that equation 5 above and equation 6 are preferably applied only to coefficient Finite part, these coefficients are included in two inverse Fourier transforms.

With reference to Fig. 6, the accurate expression of the total system time delay of asymmetric window transposer framework become as etc. Shown in formula 8:

D_ta={ (B+1) D/2-B (S_A-1) } 64/B (equation 8)

Additionally, equation 8 refers to the time delay used in the output sampling of frameworks based on 64 channel C QMF.

For using the virtual bass of asymmetric mapping window to postpone reduction system according to embodiment, Fig. 7 A shows about the first form jumping total retardation value away from size, and Fig. 7 B shows pass In second jump away from the form of total retardation value.The form 3 of Fig. 7 A shows for various conversion time delays Jump away from S for value (D=15 to 127) and the fundamental transposition factor (B=2 to 16)_AThe delay of=4. By contrast, the form 4 of Fig. 7 B shows that (D=15 is extremely for identical various change delay values 127) and jump away from S for the fundamental transposition factor (B=2 to 16)_AThe delay of=2.As can be at table Lattice 4 are seen, from the delay reduction of symmetrical 64 tap windows (D=63) to asymmetric window is 828 samplings are (for S_AFor the nominal case of=2 and B=8,2204-1376=828).

By equation 3 compared with equation 8, can verify, D is set_ts=D_taBe given:

D=L-(2 B/ (B+1)) (equation 9)

Equation 9 above is expressed the intended conversion of symmetry window as B=1 and is postponed D=L-1.

The asymmetrical amount of transposition window can change according to the restriction of system and requirement.Real in one Execute in mode and specific implementation, the two of the selected close conversion time delay of the group delay of asymmetric window/ One, in order to keep suitable transposition quality.Thus, in this case, G_d≈ D/2=20.This Can be come by the constraint of the group delay during optimizing phase of including the design of asymmetric filters Become.

Block Nyquist ptototype filter

With reference to Fig. 4, the 3rd postpones to reduce element includes that Nyquist ptototype filter 406 is blocked in use. As shown in FIG. 3 C, in order to mix virtual bass signal in hybrid domain, 8 passages and 4 lead to Road Nyquist analysis filterbank 360 is applied to virtual bass output CQMF passage (these Bank of filters is corresponding to the nyquist filter group 307 and 308 of Fig. 3 A).At a kind of embodiment In, Nyquist analysis filterbank 360 uses 13 tap ptototype filters of symmetry, and it is permissible (such as, in this case, 6 64=384 outputs are adopted to produce the time delay of 6 CQMF sampling Sample).By removing 6 coefficients acted in following sampling of ptototype filter, can be eliminated this Whole time delay (such as, 384 samplings).Usually, Nyquist analysis/synthesis chain still provides Beautiful reconstruct.But, use the frequency response of the nyquist filter group blocking wave filter to change Become.The optimization of remaining filter coefficient can improve use and block the nyquist filter of wave filter The potential poor frequency response of group.

The virtual bass signal of time lag

With reference to Fig. 4, the 4th postpones to reduce element includes making the delayed primary signal of virtual bass signal, 408.In this case, it is delayed by when broadband signal (that is, mixed signal A 364 of Fig. 3 C) During shorter compared with the time period that virtual bass system delay the practically implies that time period, can reduce The delay of whole system.Unofficially listen to test to show: do not hinder virtual low less than 20ms delayed Audio fruit.This delayed for 48kHz audio signal corresponding to 960 samplings.

In the specific implementation of embodiment, it is allowed to virtual bass signal makes broadband signal delayed altogether 352 samplings (being 7.33ms under 48kHz).Owing to 1376 can not be by CQMF bank of filters Size 64 divides exactly, so in sampling at these 352,32 samplings are to asymmetric mapping window Use.Therefore, the time delay from asymmetric window transform can be divided into the wideband delay of 1344 Bass plus 32 samplings is delayed.Thus, except 32 sampling in addition to the most delayed be 320 Individual sampling (5 CQMF samplings, corresponding to 6.67ms under 48kHz sample frequency).

The different reduction element 402 to 408 that postpones of Fig. 4 can be next with the combination of any practical quantity Use with the reduction realizing virtual bass system delay.Furthermore, it is possible to postpone minishing method by every kind Suitable change is modified to increase the delay relevant with any perception decline of virtual bass signal quality. In one embodiment, use values below achieve four postpone reduce elements: the fundamental transposition factor B=8, jumping are away from S_A=2, conversion postpones D=40, blocks nyquist filter group and 320 samplings Additional virtual bass delayed.In the illustrated case, the virtual low phonetic system in the output sampling of result System time delay is as follows:

D_VB={ (B+1) D/2-B (S_A-1) } 64/B-32+0-320=1376-352=1024

Evade the Nyquist analysis filter in pretreatment section as mentioned above (such as by using in Fig. 2 Input B 203, Fig. 3 A signal B 306 as the input in the virtual bass module 330 of Fig. 3 C D 332), the time delay of other 384 sampling can be saved, produce virtual bass system delay 1024-384=640 sampling (corresponding to 13ms under 48kHz sample frequency).

It is low that the time delay of 640 samplings in this sample situation is significantly less than previously described traditional virtual The nominal delay of 3200 samplings in system for electrical teaching.Even can be by increasing bigger virtual bass Delayed, by jumping away from S_AIncrease to 4 replacements 2, or be there is by design the result being shorter than 40 The asymmetric mapping window of analysis/synthesis time delay reduce this time delay further.But, although can Postpone with further reduction, but the change of any such value can produce the most poor virtual low Timbre matter.

Virtual bass described herein postpone reduction system embodiment can in conjunction with any suitably Virtual bass generate system system as shown in Fig. 2 and Fig. 3 use.Fig. 8 shows root Include that virtual bass generates system and postpones the audio frequency processing system of reduction system according to embodiment Block diagram.As shown in Figure 8, system 800 includes that virtual bass as shown in Figure 3 C generates system 330.Virtual low system for electrical teaching 330 receives input audio signal 801, and performs the transposition of some frequencies Function produces the audio content of enhancing with by being likely to be of raising one's voice of limited frequency response ability Device 806 plays back.Some delays can be with the transposition function phase performed by virtual low system for electrical teaching 330 Association.In one embodiment, virtual bass postpones reduction system 400 (as shown in Figure 4) It is provided as the post processing of virtual low system for electrical teaching 300 to reduce and prolonging that the process of virtual bass is associated Late.Then, the audio signal reducing delay carrying out self-virtualizing bass system 300 and 400 is sent To rendering subsystem 802, rendering subsystem 802 is configured to generate speaker feeds, and speaker is presented Send and can be fed to left and right (or multichannel) speaker 806 by amplifier 804.

Although virtual bass postpones the single post processing unit that reduction system 400 is illustrated as in system 800 Part, it should be noted that such delay reduction system may be implemented as virtual low system for electrical teaching 330 A part (as noted), or be implemented as any suitable element of system 800 A part is such as the functional part in rendering subsystem 802.Similarly, virtual low system for electrical teaching 330 is permissible It is that the traditional virtual bass summarized in background technology generates system, or it can be to use harmonic wave to change Position strengthens input audio signal 801 to increase the perception of the bass content played back by speaker 806 Any other virtual bass of level generates and processing system.

Can present by various different playback reproducers and audio tweeter (transducer) and play back Any audio frequency processing system of DAB use virtual bass postpone the embodiment of reduction system. It is arbitrary that these speakers can be presented as in the project of various different listening device or playback apparatus Kind, such as computer, TV, stereophonic sound system (family or movie theatre), cell phone, panel computer With other portable playback device.Speaker can have any suitable size and rated power, and And can driver, loudspeaker enclosure, ambiophonic system, bar shaped audio amplifier, earphone, earplug in a free-standing Deng form be provided.Speaker can be configured with any suitable array, and can include list Sound channel driver, ears speaker, surround sound loudspeaker array or any other suitable audio driven Device array.

Can include performing one or more computer of software instruction or processing means to sound Frequently signal realizes one described herein in being processed for the audio system transmitted by network Or the various aspects of more kinds of embodiment.The described any embodiment in embodiment is permissible It is used alone or be combined with each other with any combination and used.Although various embodiments are by saying The various defects of one or more local prior art discussed or mention of bright book promote, but It it is embodiment any defect of not necessarily solving in these defects.In other words, different enforcement Mode can solve the different defect that may discuss in the description.Some embodiments can only portion Point ground solves some defects that will discuss in the description or only one defect, and some embodiment party Formula can not solve any defect in these defects.

Can be at the suitable computer based sound for processing numeral or digitized audio document Process the aspect realizing system described herein in network environment.Each of adaptive audio system Part can include one or more network following: described network includes the independent of any desired quantity Machine, including for one or more router of the data of transmission between caching and route computer (not shown).Such network can build based on various different procotols, and permissible It is the Internet, wide area network (WAN), LAN (LAN) or its combination in any.

Can be by the computer that the execution of the calculating device based on processor of system is controlled Program realizes one or more in parts, block, processor or other functional part.Also should Note, according to its behavior, register transfer, logical block and/or further feature, it is possible to use hard Part, firmware and/or data and/or realization in various machine readable medias or computer-readable medium Any amount of combination of instruction describes various function disclosed herein.Such lattice can be embodied The computer-readable medium of formula data and/or instruction includes but not limited to that various forms of physics is (non- Transient state), non-volatile media, such as light, magnetic or semiconductor storage medium.

Unless the context clearly requires otherwise, otherwise throughout specification and claims, word " includes (comprise) ", " including (comprising) " etc. will be with the meaning with exclusive meaning or limit The contrary meaning included is explained；That is, explain according to the meaning of " including but not limited to ".This Outward, the word of odd number or plural number is used to respectively further comprise plural number or odd number.It addition, word is " herein In ", " hereinafter ", " above ", the word of " below " and the similar meaning refer to this Shen Any specific part of whole rather than the application please.When reference two or the list of more make During by word "or", all following explanation of this word contained in this word: in list any one The combination in any of project in all items and list in mesh, list.

Although describing one or more realization by example and according to specific embodiment, but It is to should be appreciated that one or more is practiced without limitation to disclosed embodiment.On the contrary, as to this Skilled person is it will be evident that it is intended to cover various amendment and similar layout.Therefore, appended power The scope that profit requires should meet broadest interpretation, to include all such amendments and similar cloth Put.

Claims (22)

1. for the method generating the virtual bass of low latency, including:

Receive input audio signal；

The second order using each low frequency component in the low frequency component with described input audio signal changes Described low frequency component is performed humorous by the combination transposer of position device and one or more higher order transposer Ripple replaces, to generate the transposition data of the harmonic wave representing described input audio signal；

Virtual bass signal is generated in response to described transposition data；And

By the time delay version of described virtual bass signal with described input audio signal is combined Generating the audio signal of enhancing, wherein, described harmonic transposition utilizes and uses the fundamental transposition higher than 2 The combination transposition of rank B so that described harmonic wave includes the second harmonic and extremely of each described low frequency component A few higher order harmonic wave, and make in response to by the common time using asymmetric analysis window The frequency domain value that determines to frequency domain transformation section and by the common frequency using asymmetric synthesis window The follow-up inverse transformation determined to time domain transforming section is to generate all of described harmonic wave.

Method the most according to claim 1, wherein, in response to by the common time to frequency The frequency domain value that territory transforming section determines is to generate described harmonic wave, and the described common time is to frequency domain transformation section Carry out common frequency afterwards to convert to time domain.

Method the most according to claim 1, wherein, described input audio signal is to represent to come From the threshold sampling of one group of complex values quadrature mirror filter (CQMF) subband signal or close to facing The subband CQMF signal of the low frequency audio frequency of boundary's sampling.

Method the most according to claim 3, wherein, described threshold sampling or close to critical The low frequency input audio frequency of sampling is to represent the lowest band from one group of CQMF subband signal CQMF passage 0 signal.

Method the most according to claim 4, also includes:

According to low frequency component by the following transposition data that generate: by generating asymmetric windowing zero padding Sampling and execution time to frequency domain transformation of sampling described asymmetric windowing zero padding come described Input audio signal performs the conversion of frequency domain over-sampling；And it is follow-up to from the described time to frequency domain change The output changed performs nonlinear operation to generate described transposition data according to described low frequency component；

According to the frequency component processed by described nonlinear operation by be divided in the first frequency band The second class frequency component in one class frequency component and the second frequency band generates two class frequency components；And

Further described first class frequency component is performed first frequency to time domain conversion and to institute State second class frequency component perform second frequency to time domain convert, wherein, described first frequency to time Between territory conversion and described second frequency to time domain convert in the transform size of each be described time Between to the 1/B of transform size of frequency domain transformation；And

Further asymmetric zero padding window is applied in the sampling converted from described frequency to time domain, Wherein, described asymmetric zero padding window be according to described input audio signal generate described asymmetric The 1/B length of windowing zero padding sampling, thus form two groups of transposition data.

Method the most according to claim 5, wherein, described first frequency band is from one group The frequency band of the CQMF passage 0 of CQMF subband signal, and described second frequency band is from described The frequency band of the CQMF passage 1 of one group of CQMF subband signal.

Method the most according to claim 6, wherein, generates in response to described transposition data Virtual bass signal includes being applied to one of described two groups of transposition data or both analysis filtering Device group, wherein, described analysis filterbank includes the truncated version of balanced-filter.

Method the most according to claim 7, wherein, described analysis filterbank is Nai Kuisi Special bank of filters, and the truncated version of described balanced-filter is the half symmetrical portion of described wave filter One of removed wave filter.

Method the most according to claim 8, wherein, described analysis filterbank includes that 8 lead to One of road nyquist filter group or 4 passage nyquist filter groups, and wherein, described filter One of half portion of removed symmetry of ripple device includes 6 coefficients.

Method the most according to claim 1, wherein, the time delay version of described input audio signal Originally predetermined amount of time shorter compared with the delay of described virtual bass signal it has been delayed by, and described The delayed virtual bass signal of audio signal express time strengthened.

11. methods according to claim 10, wherein, described predetermined amount of time is selected from 0 Individual sampling is to the value of scopes of 1000 samplings.

12. methods according to claim 4, wherein, directly from pretreatment hybrid filter-bank The analysis CQMF group outfan of section receives input audio frequency CQMF passage 0, thus by described pre-place The Nyquist analysis filterbank bypass of reason hybrid filter-bank section.

13. 1 kinds of equipment being used for generating the virtual bass of low latency, including:

First component, described first component receives input audio signal, and use has described input The second order transposer of each low frequency component in the low frequency component of audio signal is with one or more more The combination transposer of high-order transposer performs harmonic transposition to described low frequency component, represents described to generate The transposition data of the harmonic wave of input audio signal；And

Second component, described second component generates virtual bass signal also in response to described transposition data And be combined generating by the time delay version of described virtual bass signal Yu described input audio signal The audio signal strengthened, wherein, described harmonic transposition utilizes and uses the fundamental transposition rank B higher than 2 Combination transposition so that described harmonic wave include the second harmonic of each described low frequency component and at least one Higher order harmonic wave, and make in response to by using the common time of asymmetric analysis window to frequency Frequency domain value that territory transforming section determines and by using the common frequency of asymmetric synthesis window to the time The follow-up inverse transformation that territory transforming section determines is to generate all of described harmonic wave.

14. equipment according to claim 13, wherein, in response to by the common time to frequency The frequency domain value that rate territory transforming section determines is to generate described harmonic wave, and the described common time is to frequency domain transformation Carry out common frequency after Duan to convert to time domain.

15. equipment according to claim 13, wherein, described input audio signal is to represent From one group of complex values quadrature mirror filter (CQMF) subband signal threshold sampling or close The subband CQMF signal of the low frequency audio frequency of threshold sampling.

16. equipment according to claim 15, wherein, described threshold sampling or close to facing The low frequency audio frequency of boundary's sampling is the CQMF representing the lowest band from one group of CQMF subband signal Passage 0 signal.

17. equipment according to claim 16, also include:

3rd parts, described 3rd parts generate transposition data according to low frequency component by following: logical Cross generate asymmetric windowing zero padding sampling and to described asymmetric windowing zero padding sampling perform time Between to frequency domain transformation come to described input audio signal perform frequency domain over-sampling conversion；And it is follow-up right From the described time to the output of frequency domain transformation execution nonlinear operation with according to described low frequency component Generate described transposition data；

4th parts, described 4th parts pass through according to the frequency component processed by described nonlinear operation The second class frequency component in the first class frequency component being divided in the first frequency band and the second frequency band comes Generate two class frequency components；

5th parts, described 5th parts perform first frequency to described first class frequency component further Convert to time domain to time domain conversion and to described second class frequency component execution second frequency, its In, the conversion of described first frequency to time domain and each in the conversion of described second frequency to time domain Transform size be the described time to the 1/B of the transform size of frequency domain transformation；And

6th parts, the sampling application to converting from described frequency to time domain of described 6th parts is non- Symmetric Zero fills window, and wherein, described asymmetric zero padding window is according to described input audio signal The 1/B length of the described asymmetric windowing zero padding sampling generated, thus form two groups of transposition data.

18. equipment according to claim 17, wherein, described first frequency band is from one group The frequency band of the CQMF passage 0 of CQMF subband signal, described second frequency band is from described one group The frequency band of the CQMF passage 1 of CQMF subband signal, and wherein, in response to described transposition number According to generate virtual bass signal include being applied to one of described two groups of transposition data or both Analysis filterbank, wherein, described analysis filterbank includes the truncated version of balanced-filter.

19. equipment according to claim 18, wherein, described analysis filterbank is Nai Kui This special bank of filters, and the truncated version of described balanced-filter is symmetrical half of described wave filter The removed wave filter in one of portion.

20. equipment according to claim 19, wherein, described analysis filterbank includes 8 One of passage nyquist filter group or 4 passage nyquist filter groups, and wherein, described One of half portion of removed symmetry of wave filter includes 6 coefficients.

21. equipment according to claim 13, also include:

Timing part, being delayed by of the described timing part described audio signal of generation is virtual low with described The version of shorter predetermined amount of time is compared in the delay of tone signal；And

Hydrid component, described hydrid component is by described virtual bass signal and the described input sound through time delay Frequently signal combination is to generate the audio signal of the enhancing of the delayed virtual bass signal of express time.

22. equipment according to claim 16, also include interface unit, and described interface unit is direct Described CQMF passage is received from the analysis CQMF group outfan of pretreatment hybrid filter-bank section 0, thus the Nyquist analysis filterbank of described pretreatment hybrid filter-bank section is bypassed.